Skip to content

raia Command · Section

AI Pipeline

Convert to Markdown, train vector store, process documents.

The AI Pipeline turns raw documents into agent-usable knowledge. Files flow through Markdown conversion, chunking, embedding, and indexing into the vector store — all observable in the pipeline view.

In this section

  • See where a file is in the ingest pipeline at any moment.
  • Re-run individual steps without re-uploading.
  • Diagnose failed conversions and re-queue them.
01

What's inside

  1. 01

    Convert to Markdown

    PDF, DOCX, XLSX, PPTX, and HTML normalized into clean Markdown.

  2. 02

    Train Vector Store

    Chunk, embed, and index with model-aware token budgets.

  3. 03

    Process documents

    Optional enrichment: summarization, entity extraction, custom metadata.

Note

Re-training a vector store does not break live conversations — the swap is atomic.